CN108198546B - Voice signal preprocessing method based on cochlear nonlinear dynamics mechanism - Google Patents
Voice signal preprocessing method based on cochlear nonlinear dynamics mechanism Download PDFInfo
- Publication number
- CN108198546B CN108198546B CN201711469953.2A CN201711469953A CN108198546B CN 108198546 B CN108198546 B CN 108198546B CN 201711469953 A CN201711469953 A CN 201711469953A CN 108198546 B CN108198546 B CN 108198546B
- Authority
- CN
- China
- Prior art keywords
- nonlinear
- cochlear
- active simulation
- signal
- simulation module
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 27
- 238000007781 pre-processing Methods 0.000 title claims abstract description 21
- 238000005312 nonlinear dynamic Methods 0.000 title claims abstract description 16
- 230000007246 mechanism Effects 0.000 title claims abstract description 12
- 238000004088 simulation Methods 0.000 claims abstract description 57
- 230000004044 response Effects 0.000 claims abstract description 30
- 238000012545 processing Methods 0.000 claims abstract description 29
- 210000003477 cochlea Anatomy 0.000 claims abstract description 23
- 230000003044 adaptive effect Effects 0.000 claims description 10
- 230000003321 amplification Effects 0.000 claims description 7
- 230000000694 effects Effects 0.000 claims description 7
- 210000002768 hair cell Anatomy 0.000 claims description 7
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 7
- 238000013016 damping Methods 0.000 claims description 4
- 210000002469 basement membrane Anatomy 0.000 claims description 3
- 238000006073 displacement reaction Methods 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 abstract description 9
- 230000000737 periodic effect Effects 0.000 abstract description 6
- 230000008569 process Effects 0.000 description 5
- 238000002474 experimental method Methods 0.000 description 4
- 238000001228 spectrum Methods 0.000 description 4
- 210000000721 basilar membrane Anatomy 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 230000001629 suppression Effects 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 239000012528 membrane Substances 0.000 description 2
- 230000009022 nonlinear effect Effects 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000001886 ciliary effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 239000012530 fluid Substances 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 210000004379 membrane Anatomy 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 230000005236 sound signal Effects 0.000 description 1
- 238000010183 spectrum analysis Methods 0.000 description 1
- 230000000638 stimulation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/02—Feature extraction for speech recognition; Selection of recognition unit
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/20—Speech recognition techniques specially adapted for robustness in adverse environments, e.g. in noise, of stress induced speech
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
- G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
Landscapes
- Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Quality & Reliability (AREA)
- Signal Processing (AREA)
- Soundproofing, Sound Blocking, And Sound Damping (AREA)
Abstract
本发明公开了一种基于耳蜗非线性动力学机理的语音信号预处理方法，包括(1)建立耳蜗非线性动力学模型；(2)构建非线性耳蜗阵列；非线性耳蜗阵列为一组包含n个不同固有频率的主动仿真模块，每一个主动仿真模块将接收的输入语音信号按照耳蜗非线性动力学模型进行相应的运算后，获得每一个主动仿真模块的实时响应输出信号；(3)将各个主动仿真模块的实时响应输出信号进行处理后获得语音预处理信号。本发明由于引入了利用耳蜗非线性动力学模型的非线性耳蜗阵列代替传统的被动滤波器组对语音信号进行预处理，使得经过预处理的周期性或准周期性语音信号得到放大，并展现出与音调相关的结合音，从而提高语音处理的抗噪声能力和特征分析能力。
The invention discloses a speech signal preprocessing method based on the nonlinear dynamics mechanism of the cochlea, including (1) establishing a nonlinear dynamics model of the cochlea; (2) building a nonlinear cochlear array; the nonlinear cochlear array is a group consisting of n There are two active simulation modules with different natural frequencies. After each active simulation module performs corresponding operations on the received input speech signal according to the cochlear nonlinear dynamic model, the real-time response output signal of each active simulation module is obtained; The real-time response output signal of the active simulation module is processed to obtain a speech preprocessing signal. In the present invention, the nonlinear cochlear array using the nonlinear dynamic model of the cochlea is introduced to preprocess the speech signal instead of the traditional passive filter bank, so that the preprocessed periodic or quasi-periodic speech signal is amplified, and shows The combined tone related to the pitch, thereby improving the anti-noise ability and feature analysis ability of speech processing.
Description
Technical Field
The invention belongs to the technical field of signal processing, and particularly relates to a voice signal preprocessing method for voice recognition.
Background
The speech signal processing is the essence of modern information processing, and speech recognition and man-machine speech communication are realized through a computer. With the development of artificial intelligence technology, the level of computer speech recognition is quite high, but there is still a gap compared to humans. The problem of machine speech processing is mainly manifested in that its speech recognition capability in real scenes is susceptible to interference from ambient noise and other sound sources.
The machine voice signal processing flow mainly comprises: the method comprises the steps of voice signal preprocessing, voice signal feature extraction and voice recognition according to voice features. Artificial intelligence techniques such as neural networks, deep learning, etc. are used at the back end of the process: speech is recognized based on the extracted speech features. The speech preprocessing and speech feature extraction at the front end of the process have been implemented in the past by mathematical signal processing methods, such as: fourier analysis, wavelet transform, etc. To further improve the capability of machine speech recognition, people are increasingly inclined to use auditory signal processing mechanisms for speech signal processing. Existing hearing-based speech signal processing methods use band-pass filter banks with different center frequencies to simulate the frequency analysis function of the cochlea. The gamma tone is most commonly used as the impulse response of the filter. These hearing-based signal processing provide some improvement to speech signal processing, but the existing cochlear filter model, as a linear model, is very different from the real cochlea.
Disclosure of Invention
Aiming at the defects of the prior art, the invention aims to provide a voice signal preprocessing method based on a cochlear nonlinear dynamics mechanism, and aims to further improve the feature analysis and the anti-noise interference capability of the existing machine voice signal processing technology.
The invention provides a voice signal preprocessing method based on a cochlear nonlinear dynamics mechanism, which comprises the following steps:
(1) establishing a cochlear nonlinear dynamical model;
(2) constructing a nonlinear cochlear array according to a cochlear nonlinear dynamical model;
the nonlinear cochlear array is a group of active simulation modules containing n different natural frequencies, and each active simulation module performs corresponding operation on the received input voice signal according to the cochlear nonlinear dynamics model to obtain a real-time response output signal of each active simulation module;
(3) processing the real-time response output signals of the active simulation modules to obtain voice preprocessing signals;
wherein n is the number of the active simulation modules, and n is an integer greater than or equal to 1.
Further, the nonlinear dynamical model of cochlea is:
wherein x is the basement membrane off-balancePosition displacement, t is time, gamma is damping coefficient, gammaαIs the adaptive force coefficient, B is the outer hair cell electrostrictive coefficient, x0Is the primary length of outer hair cells, omegaiIs the natural circular frequency of the cochlea, S (t) is the input speech signal, xiAnd (t) is a real-time response output signal of the ith active simulation module, i is a serial number of the active simulation module, and i is 1,2,3.
Further, the adaptive force coefficient gamma in the cochlear nonlinear dynamics modelαThe following ranges should be satisfied: gamma is more than 0αGamma is less than or equal to gamma, within this range gammaαThe larger the value, the greater the amplification of the speech signal near its natural frequency by the active simulation module.
Further, the natural frequencies of the n active simulation modules can be set as follows: for natural frequency range a-a × eε(n-1)Hz (epsilon < 1) nonlinear cochlear array, wherein the natural frequency of the ith active simulation module is fi＝a*eε(i-1)Hz; i is the serial number of the active simulation module, i 1,2,3.
Wherein a is more than or equal to 20Hz and less than or equal to 200 Hz.
Further, in step (3), the real-time response output signal x is averaged according to an energy methodi(T) processing to obtain speech frame signal with time length T
Wherein x isi(t) is the real-time response output signal of the ith active simulation module, yiAnd (T) is the signal after pretreatment, T is time, and T is the duration of a voice frame.
Compared with the prior art, the technical scheme of the invention has the advantages that the nonlinear cochlear array utilizing the cochlear nonlinear dynamical model is introduced to replace the traditional passive filter bank to preprocess the voice signal, so that the preprocessed periodic or quasi-periodic voice signal is amplified, and the combined voice related to the tone is displayed, thereby improving the anti-noise capability and the characteristic analysis capability of the voice processing.
Drawings
Fig. 1 is a technical block diagram of the present invention for implementing a novel speech signal processing by using a cochlear nonlinear dynamical model.
FIG. 2(a) is a response spectrum of a nonlinear cochlear model; fig. 2(b) shows a response spectrum in a cochlear physiological experiment.
FIG. 3(a) is a spectrogram of a speech after adding noise; fig. 3(b) is a spectrogram processed by the active simulation module.
FIG. 4 is a graph comparing frequency response characteristics of an active simulation module and a passive system. In the figure, the solid line is the frequency response characteristic line of the active simulation module, and the dotted line is the frequency response characteristic line of the passive system.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The cochlea is a nonlinear signal processing system, and has the characteristics of two-tone suppression, generation of combined tones and the like, and the characteristics play an important role in signal processing. For example, the term "combined tone" means that when two frequencies are f1、f2When excited by sound of (2), there will be pf1+mf2Frequency component components of the excitation signal (p, m are integers) appear. Wherein the difference frequency component f1-f2The composite sound without fundamental frequency and with higher harmonic passes through cochlea to generate corresponding fundamental frequency, so that the tones with fundamental frequency which does not exist still exist. The cochlea also enables the tones of the quasi-periodic signal to be perceived through nonlinearity. The nonlinear characteristic of the cochlea can be described by a nonlinear dynamical equation, and the invention designs an active simulation module capable of completing the mathematical operation according to the nonlinear dynamical equation and according to the parameter omega in the equationiThe nonlinear cochlear array formed by n active simulation modules with different natural frequencies is designed for preprocessing voice by different values (representing the natural circular frequencies of the active simulation modules). Compared with the conventional method using band-pass filter bankCompared with the sound processing strategy, the voice signal processed by the nonlinear cochlear array can reflect a plurality of nonlinear effects related to the sound processing mechanism of the cochlea, such as nonlinear tuning, polyphonic distortion, two-tone suppression and the like similar to the auditory processing result, particularly can show a combination tone related to the tone, and an active amplification mechanism of a periodic or quasi-periodic voice signal close to the natural frequency of an active simulation module, so that the voice processing has better characteristic analysis and anti-noise capability than the traditional method.
In order to improve the speech recognition and anti-noise interference capability of the existing machine speech signal processing in the real environment. The invention constructs a voice signal preprocessing method based on a cochlear nonlinear dynamics model. The method introduces a nonlinear cochlear array utilizing a cochlear nonlinear dynamical model to replace a traditional passive filter bank to preprocess voice signals. Analysis shows that the simulation result of the nonlinear cochlear array is highly consistent with the physiological experiment result of the cochlear basilar membrane and the auditory psychological experiment result, and especially, a plurality of nonlinear effects related to the sound processing mechanism of the cochlea, such as nonlinear tuning, polyphonic distortion, two-tone suppression and the like, can be well simulated. By using the method to process the voice signal, the combined voice related to the tone can be displayed, the periodic signal characteristic related to the voice is enhanced, and the voice signal is highlighted from the noise, so that the voice identification degree is improved.
The invention adopts the following specific technical scheme:
(1) establishing a cochlea nonlinear dynamics model:
we take cochlear basilar membrane dynamics as the basis, and take cochlear local as an example for stress analysis. During the acoustic conduction process, the cochlea basal membrane is locally subjected to external force F caused by external acoustic stimulations(t), base film self-elasticity FTLymphatic fluid and self-generated resistance
And the nonlinear adaptive force F regulated by the electrostriction of the outer hair cells and the ciliary movementaIts simplified expression is as follows:
the nonlinear dynamical model of the cochlea built according to the Newton's law of mechanics is as follows:
wherein x is the displacement of the basement membrane from the equilibrium position, gamma is the damping coefficient, gammaαIs the adaptive force coefficient, B is the outer hair cell electrostrictive coefficient, x0Is the primary length of outer hair cells, omegaiThe input signal is S (t) which is the natural circular frequency of the cochlea in this region. Solving the nonlinear equation to obtain the real-time response output x of the cochlear basilar membranei(t)。
(2) A voice signal preprocessing method added with a cochlear nonlinear dynamics model comprises the following steps:
as shown in fig. 1, the new speech signal preprocessing method requires constructing a nonlinear cochlear array according to a nonlinear dynamical model of the cochlea to simulate the processing mechanism of the cochlea on sound. The nonlinear cochlear array is a group of n cochlear nonlinear dynamical models with different natural frequencies
The active simulation module of (2) to form a nonlinear simulation array. The input speech signal is S (t), the real-time response output x of different processed channels can be obtained by solving the equationi(t) of (d). Then each channel outputs a signal xi(t) averaging according to an energy method to obtain a speech preprocessing signal
It should be noted that the design of the active simulation module should be such that the adaptive force coefficient γ isαThe settings were: gamma is more than 0αIn the range of ≦ γ, when γ is presentαWhen the adaptive force is 0, the system becomes a passive system; when gamma isαWhen γ, the system will eventually oscillate self-sustained. When gamma isαWithin the above range, γαThe larger the value is, the larger the maximum value of the adaptive force is, and the active simulation isThe smaller the effective damping of the module, the greater the amplitude of the response to a speech signal near the natural frequency of the active simulation module. The frequency response curve of the active simulation module and the frequency response characteristic curve of the passive system are shown in fig. 4, for example, it can be seen that the active simulation module has a better amplification effect on the voice signal near its natural frequency.
The present invention will be described in further detail with reference to the accompanying drawings and specific examples.
Fig. 1 is a block diagram of a novel speech signal processing technique implemented by using a cochlear nonlinear dynamical model. The specific strategy is as follows: n frequency band channels are required to be designed according to the frequency, and each frequency band channel comprises an active simulation module with different natural frequencies to form a nonlinear cochlear array. The sound signal S (t) recorded by the microphone is processed and output as x by different simulation modulesi(t), thereafter xi(t) average output by energy of yi(t)，yiAnd (t) is the signal after pretreatment. The preprocessing signal obtained by the method can reflect the combined tone information which is consistent with the cochlear processing result and is related to auditory tones, and enhance the periodic signal characteristics related to voice, so that the voice signal characteristics are highlighted in noise, and the voice recognition degree is improved.
Fig. 2 shows the multi-tone distortion effect, and as can be seen from the response spectrum of the active simulation module in fig. 2(a), distortion products which cannot be found in the conventional passive filter system appear in the response spectrum of the nonlinear cochlear array, namely: and combining the sound. Fig. 2(b) is the response result on the cochlear basilar membrane of the actual physiological experiment. The comparison shows that the active simulation module can well simulate the polyphonic distortion effect in the cochlea, and combined tone information related to the voice tones appears in the voice signals processed by the active simulation module, and meanwhile, the fundamental frequency signals of the voice are improved, which is also the basis for improving the voice characteristics by the strategy.
Fig. 3(a) is a result of the richness spectrum analysis of a speech after adding noise, and it can be seen from the figure that the speech features are almost drowned by the noise. The results of the Fourier transform analysis using the model constructed in this study are shown in FIG. 3 (b). It can be seen that speech signal features are clearly highlighted in the noise.
FIG. 4 is a graph comparing frequency response characteristics of an active simulation module and a passive system. The horizontal axis of the image is the sound frequency, and the vertical axis is the response amplitude of the system to the sounds with different frequencies. In the figure, the natural frequencies of the active simulation module and the passive system are both 140Hz, the solid line is the frequency response characteristic line of the active simulation module, and the dotted line is the frequency response characteristic line of the passive system. Compared with a passive system, the active simulation module has larger response amplitude to the voice signals near the natural frequency, and the active amplification effect of the active simulation module on the voice signals is embodied.
The invention provides a voice signal preprocessing method based on a cochlear nonlinear dynamics mechanism, which comprises the following steps:
(1) establishing a cochlear nonlinear dynamics model
(2) And constructing a nonlinear cochlear array according to the cochlear nonlinear dynamical model, wherein the nonlinear cochlear array is a group of active simulation modules containing n different natural frequencies, and the nonlinear cochlear array is used for performing corresponding mathematical operation according to the nonlinear dynamical model. The natural frequency of the active simulation module can be set according to the following modes: e.g. for the natural frequency range a-eε(n-1)Hz (generally, 20Hz is more than or equal to a and less than or equal to 200Hz) nonlinear cochlear array, wherein the natural frequency of the ith active simulation module is fi＝a*eε(i-1)Hz (i ═ 1,2,3.. n). The design of the active simulation module should make the adaptive force coefficient gammaαThe settings were: gamma is more than 0αA range of ≦ γ, in which range γ is presentαThe larger the value is, the larger the amplification effect of the active simulation module on the voice signal near the natural frequency is;
the input voice signal is S (t), the real-time response output x of each processed active simulation module can be obtained by solving the equationi(t)；
(3) Output signal x for each channeli(t) processing to obtain a speech pre-processed signal
Wherein the channel output signals x can be processed according to energy-wise averagingi(T), thereby obtaining the voice frame signal with the time length T for the subsequent voice processing process.
The nonlinear cochlear array can align periodic or periodical voice signals to play an active amplification role, so that the voice signals required by people can be highlighted from noise. Meanwhile, the nonlinear cochlear array can well simulate the polyphonic distortion effect of the cochlea, so that the preprocessed signals can show combined tones related to tones, the voice characteristics are highlighted, and the voice recognition degree is improved.
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (3)
1. A speech signal preprocessing method based on a cochlear nonlinear dynamics mechanism is characterized by comprising the following steps:
(1) establishing a cochlear nonlinear dynamical model, wherein the cochlear nonlinear dynamical model comprises the following steps:
(2) constructing a nonlinear cochlear array according to a cochlear nonlinear dynamical model, wherein the nonlinear cochlear array is a group of active simulation modules comprising n different natural frequencies, and each active simulation module performs corresponding operation on a received input voice signal according to the cochlear nonlinear dynamical model to obtain a real-time response output signal of each active simulation module;
(3) processing the real-time response output signals of the active simulation modules to obtain voice preprocessing signals;
in the step (3), leveling is performed according to an energy methodAll respond to real-time output signal xi(T) processing to obtain speech frame signal with time length T
Wherein x is the displacement of the basement membrane from the equilibrium position, t is time, gamma is the damping coefficient, gammaαAs adaptive force coefficient, the adaptive force coefficient gamma in the cochlea nonlinear dynamical modelαThe following ranges should be satisfied: gamma is more than 0αGamma is less than or equal to gamma, within this range gammaαThe larger the value is, the larger the amplification effect of the active simulation module on the voice signal near the natural frequency is, B is the outer hair cell electrostriction coefficient, x0Is the primary length of outer hair cells, omegaiIs the natural circular frequency of a certain part of the cochlea, S (t) is the input voice signal, xi(t) is a real-time response output signal of the ith active simulation module, i is a serial number of the active simulation module, and i is 1,2,3i(T) is the signal after pretreatment, and T is the duration of a voice frame; n is the number of the active simulation modules, and n is an integer greater than or equal to 1.
2. The speech signal preprocessing method of claim 1 wherein the natural frequencies of the n active simulation modules are set as follows:
for natural frequency range a-a × eε(n-1)Hz non-linear cochlear array, wherein the i-th active simulation module has a natural frequency fi＝a*eε(i-1)Hz；ε＜1。
3. The speech signal preprocessing method of claim 2 wherein 20Hz ≦ a ≦ 200 Hz.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711469953.2A CN108198546B (en) | 2017-12-29 | 2017-12-29 | Voice signal preprocessing method based on cochlear nonlinear dynamics mechanism |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711469953.2A CN108198546B (en) | 2017-12-29 | 2017-12-29 | Voice signal preprocessing method based on cochlear nonlinear dynamics mechanism |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108198546A CN108198546A (en) | 2018-06-22 |
| CN108198546B true CN108198546B (en) | 2020-05-19 |
Family
ID=62586109
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711469953.2A Expired - Fee Related CN108198546B (en) | 2017-12-29 | 2017-12-29 | Voice signal preprocessing method based on cochlear nonlinear dynamics mechanism |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108198546B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111510837B (en) * | 2020-04-23 | 2023-05-30 | 武汉立江科学仪器有限责任公司 | Hearing recovery method, recovery system, storage medium and hearing aid |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5388182A (en) * | 1993-02-16 | 1995-02-07 | Prometheus, Inc. | Nonlinear method and apparatus for coding and decoding acoustic signals with data compression and noise suppression using cochlear filters, wavelet analysis, and irregular sampling reconstruction |
| CN100502819C (en) * | 2005-05-24 | 2009-06-24 | 北京大学科技开发部 | Artificial cochlea manufacture method suitable for Chinese voice coding strategy |
| CN101645267B (en) * | 2009-04-03 | 2012-02-01 | 中国科学院声学研究所 | A Speech Processing Method Applied to Cochlear Electronics |
| CN102947883A (en) * | 2010-01-29 | 2013-02-27 | 循环逻辑有限责任公司 | Method and apparatus for canonical nonlinear analysis of audio signals |
| CN106621033B (en) * | 2016-11-30 | 2019-09-27 | 深圳信息职业技术学院 | Cochlear Electronic Speech Processing Method and System |
-
2017
- 2017-12-29 CN CN201711469953.2A patent/CN108198546B/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN108198546A (en) | 2018-06-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109326302B (en) | Voice enhancement method based on voiceprint comparison and generation of confrontation network | |
| US11800301B2 (en) | Neural network model for cochlear mechanics and processing | |
| Cheng et al. | Replay detection using CQT-based modified group delay feature and ResNeWt network in ASVspoof 2019 | |
| CN110379412A (en) | Method, apparatus, electronic equipment and the computer readable storage medium of speech processes | |
| Rawat et al. | Emotion recognition through speech using neural network | |
| CN114613389A (en) | Non-speech audio feature extraction method based on improved MFCC | |
| Li et al. | Densely connected network with time-frequency dilated convolution for speech enhancement | |
| WO2024103485A1 (en) | Human speech structure-based speech interference noise design method | |
| CN108198546B (en) | Voice signal preprocessing method based on cochlear nonlinear dynamics mechanism | |
| CN108053829B (en) | An Electronic Cochlear Coding Method Based on Cochlear Auditory Nonlinear Dynamics | |
| CN102934158B (en) | Learning and auditory scene analysis in gradient frequency nonlinear oscillator networks | |
| CN108417198A (en) | A Method of Male-female Speech Conversion Based on Spectrum Envelope and Pitch Period | |
| Chen et al. | A study of deep belief network based Chinese speech emotion recognition | |
| JP2013518313A (en) | Method and apparatus for canonical nonlinear analysis of speech signals | |
| Aggarwal et al. | Performance evaluation of artificial neural networks for isolated Hindi digit recognition with LPC and MFCC | |
| CN113763978B (en) | Voice signal processing method, device, electronic equipment and storage medium | |
| Zheng et al. | Throat microphone speech enhancement via progressive learning of spectral mapping based on lstm-rnn | |
| CN114495908A (en) | Method and system for driving mouth shape by voice based on time sequence convolution | |
| Alam et al. | Speaker identification system under noisy conditions | |
| Torfs et al. | Artifact-free Sound Quality in DNN-based Closed-loop Systems for Audio Processing | |
| CN111816187A (en) | Deep neural network-based voice feature mapping method in complex environment | |
| Cataldo et al. | IDENTIFICATION OF A STOCHASTIC PROCESS MODELING THE STIFFNESS OF THE VOCAL FOLDS FOR A VOICE PRODUCTION MODEL REPRESENTED BY A NEURAL NETWORK (EURODYN 2020) | |
| Vimal | Study on the behaviour of Mel frequency cepstral coffecient algorithm for different windows | |
| Wan et al. | Speaker recognition based on BP neural network | |
| CN118013329A (en) | Underwater sound signal LOFAR spectrogram sample amplification method based on generation countermeasure network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20200519 Termination date: 20201229 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |